A Comprehensive Approach to Predict Kidney Disease Risk Factor

Chronic kidney disease(CKD), is a type of kidney failure or renal failure that extends over months or years, and Early diagnosis is vital to fight the disease and ensure proper treatment. This work develops a web application with user interface to prompt out whether a patient is prone to CDK or not. To necessitate this, the prominent features are selected using feature selection techniques and classification is performed by widely used classifiers. Each feature is computed by a score statistically utilizing feature Selection methods namely, Chi square, Correlation, Mutual Information, Fisher score and Mean Absolute Difference. Optimality test is performed to identify the optimal number of minimum features required and was found to be 5. Further, maximum voting technique is applied to identify the top 5 features among the results of all techniques. Accordingly, a ranking table is generated from which the top 5 features are identified as $\{$ Red blood cell count, Albumin, Packed Cell Volume, Hypertension, Specific gravity}. Different Classification Algorithms are used to predict the class labels. The performance results indicate best random forest, decision tree and logistic regression are the top 3 classifiers with accuracies of $\mathbf{9 7. 5 6 \%}, \quad \mathbf{9 5. 6 2 \%}, \mathbf{9 6. 2 5 \%}$ respectively; and thus, these classifiers were utilized for ensembling and have exhibited an accuracy of $\mathbf{9 7. 5 \%}$.